Metabolic Engineering of Escherichia coli for Enhanced Production of Succinic Acid, Based on Genome Comparison and In Silico Gene Knockout Simulation

Transcription

1 APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Dec. 2005, p Vol. 71, No /05/$ doi: /aem Copyright 2005, American Society for Microbiology. All Rights Reserved. Metabolic Engineering of Escherichia coli for Enhanced Production of Succinic Acid, Based on Genome Comparison and In Silico Gene Knockout Simulation Sang Jun Lee, 1,2 Dong-Yup Lee, 1,3 Tae Yong Kim, 1 Byung Hun Kim, 1 Jinwon Lee, 4 and Sang Yup Lee 1,2,3 * Metabolic and Biomolecular Engineering National Research Laboratory, Department of Chemical and Biomolecular Engineering, 1 Center for Ultramicrochemical Process Systems, 2 and Department of BioSystems, BioProcess Engineering Research Center and Bioinformatics Research Center, 3 Korea Advanced Institute of Science and Technology, Guseong-dong, Yuseong-gu, Daejeon , Republic of Korea, and Department of Chemical and Biomolecular Engineering, Sogang University, Seoul , Republic of Korea 4 Received 19 May 2005/Accepted 15 August 2005 Comparative analysis of the genomes of mixed-acid-fermenting Escherichia coli and succinic acid-overproducing Mannheimia succiniciproducens was carried out to identify candidate genes to be manipulated for overproducing succinic acid in E. coli. This resulted in the identification of five genes or operons, including ptsg, pykf, sdha, mqo, and aceba, which may drive metabolic fluxes away from succinic acid formation in the central metabolic pathway of E. coli. However, combinatorial disruption of these rationally selected genes did not allow enhanced succinic acid production in E. coli. Therefore, in silico metabolic analysis based on linear programming was carried out to evaluate the correlation between the maximum biomass and succinic acid production for various combinatorial knockout strains. This in silico analysis predicted that disrupting the genes for three pyruvate forming enzymes, ptsg, pykf, and pyka, allows enhanced succinic acid production. Indeed, this triple mutation increased the succinic acid production by more than sevenfold and the ratio of succinic acid to fermentation products by ninefold. It could be concluded that reducing the metabolic flux to pyruvate is crucial to achieve efficient succinic acid production in E. coli. These results suggest that the comparative genome analysis combined with in silico metabolic analysis can be an efficient way of developing strategies for strain improvement. Succinic acid is one of the fermentation products of anaerobic metabolism as well as an intermediate of the tricarboxylic acid cycle. It has been used as a precursor for various chemicals, a food additive, an ion chelator, and a supplement to pharmaceuticals (34). Succinic acid has mostly been produced chemically from maleic anhydride. Recently, fermentative production of succinic acid has been receiving much research attention, as several bacteria can produce succinic acid as a major fermentation product. The naturally isolated obligate anaerobe Anaerobiospirillum succiniciproducens (6) and facultative anaerobes belonging to the family Pasteurellaceae, such as Actinobacillus succinogenes (10) and Mannheimia succiniciproducens (16), have been shown to be able to produce succinic acid efficiently. Escherichia coli also produces succinic acid but as a minor fermentation product. E. coli prefers to produce much more acetic acid, formic acid, lactic acid, and ethanol rather than succinic acid during anaerobic fermentation. Thus, it is necessary to redirect metabolic fluxes for increasing succinic acid production as well as reducing formation of other metabolites. Toward this goal, an ldha and pfl double mutant E. coli strain * Corresponding author. Mailing address: Department of Chemical and Biomolecular Engineering, Korea Advanced Institute of Science and Technology, Guseong-dong, Yuseong-gu, Daejeon , Republic of Korea. Phone: Fax: leesy@kaist.ac.kr. These authors contributed equally to this work. NZN111 was developed to block the formation of lactic, acetic, and formic acids (2). However, cell growth was relatively slow, possibly due to the inactivation of enzymes involved in pyruvate dissimilation (3). Chatterjee and coworkers (4) reported that an additional ptsg gene mutation recovered the growth of NZN111 to some extent. However, acetic acid, formic acid, and ethanol were still formed. Amplification of CO 2 -fixing anaplerotic pathways catalyzed by phosphoenolpyruvate (PEP) carboxylase (18), malic enzyme (12, 27, 28), and pyruvate carboxylase (9, 31) have resulted in the enhancement of succinic acid production in E. coli. Recently, Kim et al. (14) reported that amplification of the Actinobacillus succinogenes PEP carboxykinase in a PEP carboxylase-negative E. coli strain increased succinic acid production. Lin et al. (17) showed that succinic acid could be aerobically produced by utilizing the glyoxylate cycle in E. coli. Also, Sanchez et al. (23) reported impressive results for anaerobic production of succinic acid with a cumulative yield of 160 mm of succinic acid from 100 mm glucose in 24 h by repeated feeding fermentation of recombinant E. coli. We recently reported the complete genome sequence of M. succiniciproducens MBEL55E, a gram-negative capnophilic bacterium, which was isolated from bovine rumen (16), and explored the genome-scale metabolic characteristics leading to high-level succinic acid production (11). It was reasoned that efficient production of succinic acid by E. coli would be possible by engineering its metabolic pathways to mimic those of M. succiniciproducens. 7880

2 VOL. 71, 2005 METABOLIC ENGINEERING FOR SUCCINIC ACID PRODUCTION 7881 TABLE 1. Bacterial strains and plasmids used in this study Strain or plasmid Description a Source or reference E. coli strains W3110 Coli Genetic Stock Center strain (CGSC) no CGSC W3110G ptsg::sp r This study W3110GF ptsg::sp r pykf::tc r This study W3110GFO ptsg::sp r pykf::tc r mqo::cm r This study W3110GFH ptsg::sp r pykf::tc r sdh::km r This study W3110GFHO ptsg::sp r pykf::tc r sdh::km r mqo::cm r This study W3110GFHOE ptsg::sp r pykf::tc r sdh::km r mqo::cm r aceba::pm r This study W3110GFA ptsg::sp r pykf::tc r pyka::km r This study W3110GFAP ptsg::sp r pykf::tc r pyka::km r pflb::cm r This study W3110GFAPL ptsg::sp r pykf::tc r pyka::km r pflb::cm r ldha::pm r This study Plasmids pkd46 Red recombinase expression plasmid, temperature-sensitive ori; Ap r 5 pacyc184 Cm r Tc r New England Biolabs b pic156 Sp r 26 puc19-phleo Ap r Pm r 7 puc4k Ap r Km r Pharmacia c a Abbreviations: Ap, ampicillin; Cm, chloramphenicol; Km, kanamycin; Pm, phleomycin; Sp, spectinomycin; Tc, tetracycline. b New England Biolabs, Inc., Beverly, Mass. c Pharmacia Biotech, Uppsala, Sweden. In this paper, we report comparative analysis of the central metabolic pathways of M. succiniciproducens and E. coli to select important metabolic pathways with respect to the formation of fermentation products. Subsequent in silico analysis was performed to predict the target genes among the candidates to be removed from E. coli for the efficient production of succinic acid. Finally, these predictions were validated by actual gene knockout experiments. In addition, the reasons for different fermentation characteristics between the rumen bacterium and E. coli are discussed. Template PCR no. TABLE 2. Oligonucleotides used for gene replacement experiments Primer Sequence (5 3 3 ) SpR 1 PTSG1 TGC CCG CCG TTG TAT CGC ATG TTA TGG CAG GGG GAT CGA TCC TCT AGA PTSG2 TGC AGC AAC CAG AGC CGG TGC CAT TTC GCT GGG CCG ACA GGC TTT 2 PTSG3 TGG GCG TCG GTT CCG CGA ATT TCA GCT GGC TGC CCG CCG TTG TAT CGC PTSG4 GAG GTT AGT AAT GTT TTC TTT ACC ACC AAA TGC AGC AAC CAG AGC CGG TcR 1 PYKF1 TGG ACG CTG GCA TGA ACG TTA TGC GTC TGA GGG TAG ATT TCA GTG CAA PYKF2 CGC CTT TGC TCA GTA CCA ACT GAT GAG CCG GGG TTC CAT TCA GGT CGA 2 PYKF3 CCG AAT CTG AAG AGA TGT TAG CTA AAA TGC TGG ACG CTG GCA TGA ACG PYKF4 AAG TGA TCT CTT TAA CAA GCT GCG GCA CAA CGC CTT TGC TCA GTA CCA CmR 1 MQO1 GGC ATA CCA TGC CGG ATG TGG CGT ATC ATT GGG GTT TAA GGG CAC CAA MQO2 GAA CTA CGG CGA GAT CAC CCG CCA GTT AAT GCC CCG GGC TTT GCG CCG 2 MQO3 TGG CGC GTC TTA TCA GCA TAC GCC ACA TCC GGC ATA CCA TGC CGG ATG MQO4 AGC CAC GCG TAC GGA AAT TGG TAC CGA TGT GAA CTA CGG CGA GAT CAC KmR 1 SDH1 CAG TCA GAG AAT TTG ATG CAG TTG TGA TTG ATC GGG GGG GGG GGA AAG SDH2 ATC GGC TCT TTC ACC GGA TCG ACG TGA GCG ATC CCA ATT CTG ATT AGA 2 SDH3 GTT GTG GTG TGG GGT GTG TGA TGA AAT TGC CAG TCA GAG AAT TTG ATG SDH4 ATC ATG TAG TGA CAG GTT GGG ATA ACC GGA ATC GGC TCT TTC ACC GGA PmR 1 ACEBA1 GCA CCT TGT GAT GGT GAA CGC ACC GAA GAA CGA GCT CGG TAC CCG GGC ACEBA2 CTT TCG CCT GTT GCA GCG CCT GAC CGC CAG CAA TAG ACC AGT TGC AAT 2 ACEBA3 GAC GCG CCG ATT ACT GCC GAT CAG CTG CTG GCA CCT TGT GAT GGT GAA ACEBA4 ATC CCG ACA GAT AGA CTG CTT CAA TAC CCG CTT TCG CCT GTT GCA GCG KmR 1 PYKA1 CAC CTG GTT GTT TCA GTC AAC GGA GTA TTA CAT CGG GGG GGG GGG AAA G PYKA2 GTG GCG TTT TCG CCG CAT CCG GCA ACG TAC ATC CCA ATT CTG ATT AG 2 PYKA3 TTA TTT CAT TCG GAT TTC ATG TTC AAG CAA CAC CTG GTT GTT TCA GTC PYKA4 GTT GAA CTA TCA TTG AAC TGT AGG CCG GAT GTG GCG TTT TCG CCG CAT C PmR 1 LDHA1 CGG TGT GGC GAT GCT GCG CAT TCT GAA AGG CAA TAG ACC AGT TGC AAT LDHA2 GTC AGA GCT TCT GCT GTC AGG AAT GCC TGG CGA GCT CGG TAC CCG GGC 2 LDHA3 AAC GGC AGG CGT TAT CGG TAC CGG TAA AAT CGG TGT GGC GAT GCT GCG LDHA4 AAG TTT TGC AGC GTA GTC TGA GAA ATA CTG GTC AGA GCT TCT GCT GTC CmR 1 PFLB1 GGG GTG TTA CCC GTT TTC TTA CCA TAC ACA GGG GTT TAA GGG CAC CAA PFLB2 AGT GGA CTT ACT TCG GCT ACC TGG CTG CTG GGG CTT TGC GCC GAA TAA 2 PFLB3 CCG AAC GGC GCG CCA GCA CGA CGA CCG TCT GGG GTG TTA CCC GTT TTC PFLB4 GTC CGG CTA CCA ACG CTC AGG AAG CTA TCC AGT GGA CTT ACT TCG GCT

3 7882 LEE ET AL. APPL. ENVIRON. MICROBIOL. FIG. 1. Comparison of the central metabolic pathways related to succinic acid formation in M. succiniciproducens (A) and E. coli (B). Gene names are shown in italics. Dashed arrows in panel B represent those metabolic pathways only existing in E. coli. Acetyl-coenzyme A (Acetyl-CoA) synthetase (acs), fumarase AB (fumab), glyoxylate shunt (aceba), malate dehydrogenase (mqo), NAD-dependent malic enzyme (sfca), phosphotransferase system (ptsg), pyruvate kinase F (pykf), pyruvate oxidase (poxb), and succinate dehydrogenase (sdhabcd) genes exist in E. coli but not in M. succiniciproducens. MATERIALS AND METHODS Strains and plasmids. All strains and plasmids used in this study are listed in Table 1. All DNA manipulations were carried out by following standard protocols (22). Plasmid pkd46 (5), harboring the bacteriophage red operon, which contains the exo, beta, and gam genes, was used for disrupting the genes of interest in the chromosome of E. coli W3110 with appropriate antibiotic markers (Table 2). Recombinant E. coli W3110 harboring pkd46 was cultivated at 30 C, and recombinases were induced by adding L-arabinose (1 mm) at an optical density at 600 nm (OD 600 ) of 0.4. Then electrocompetent cells were prepared by standard protocol (22). Two-step PCRs were performed using the antibiotic resistance genes as templates and primers listed in Table 2. The PCR products were transformed into the electrocompetent E. coli W3110 harboring pkd46. Colonies were selected on Luria-Bertani (LB; 10 g tryptone liter 1, 5 g yeast extract liter 1, 10 g NaCl liter 1 ) agar plates containing the appropriate antibiotics at the following concentrations ( gml 1 ): chloramphenicol, 34; kanamycin, 25; phleomycin, 5; tetracycline, 15; spectinomycin, 50. Successful gene replacement with the antibiotic marker was confirmed by PCR. Comparative genome analysis. The central metabolic pathways of E. coli K-12 were reconstructed using the information extracted from the metabolic database BioSilico (13). The fermentation pathways of M. succiniciproducens MBEL55E were reconstructed based on the recent genome annotation ( Two reconstructed metabolic pathways were compared to select the target genes in E. coli to be manipulated for mimicking efficient succinic acid production in M. succiniciproducens. Glucose phosphorylation assay. To elucidate whether M. succiniciproducens uses a PEP-sugar phosphotransferase system (PTS) to actively uptake glucose, 14 C-labeled glucose phosphorylation assay was carried out as described previously (25). Cell extracts of E. coli W3110 and M. succiniciproducens MBEL55E were employed for the assays using 14 C-labeled glucose as a substrate and PEP or ATP as a phosphate donor. The amounts of phosphorylated glucose passed through a polystyrene column were analyzed by a scintillation counter. In silico analysis. A genome-scale in silico E. coli model (8, 21) was used with slight modifications. The specific growth rate can be calculated by a biomass equation derived from the drain of biosynthetic precursors into E. coli biomass with their appropriate ratios (19). Under the pseudo-steady-state assumption, the unknown internal fluxes within the underdetermined metabolic reaction network can be evaluated by means of linear programming subject to the constraints provided by mass conservation, thermodynamics, and reaction stoichiometry (8, 24, 30). The in silico analysis was carried out using the MetaFluxNet program package, version 1.69 (15). Culture conditions. Cells were routinely cultured in LB medium at 37 C. Antibiotics were added at the concentrations shown above depending on the knockout markers used. For the fermentation experiments, cells were first grown in 10 ml LB medium at 37 C with shaking. One milliliter of seed culture was used to inoculate a 125-ml butyl rubber-stoppered serum vial containing 100 ml of fermentation medium, which contains (per liter): glucose, 9 g (50 mm); yeast extract, 5 g; NaHCO 3,10 g; NaH 2 PO 4 H 2 O, 8.5 g; K 2 HPO 4,15.5 g (ph 7.0). The vial headspace was filled with CO 2, and Na 2 S 9H 2 O was added to a final concentration of 1 mg liter 1 to ensure anaerobic conditions. Analytical procedure. For the measurement of the concentrations of glucose and organic acids, culture supernatant was passed through a syringe filter (pore size of 0.2 m) after centrifugation at 10,000 g for 10 min. The glucose concentration was determined by using a glucose analyzer (model 2300 STAT; Yellow Springs Instrument Co., Yellow Springs, Ohio). The concentrations of organic acids were determined by high-performance liquid chromatography (L RI monitor, L-4200 UV-VIS detector, D2500 chromato-integrator; Hitachi, Tokyo, Japan) equipped with an Aminex HPX-87H column (300 by 7.8 mm, Hercules, Calif.). The column was eluted isocratically at 50 C at a flow rate of 0.6 ml/min with 0.01 N H 2 SO 4. Cell growth was monitored by measuring the absorbance at 600 nm (OD 600 ) using an Ultrospec3000 spectrophotometer (Pharmacia Biotech, Uppsala, Sweden). RESULTS Comparison of central metabolic pathways of E. coli and M. succiniciproducens. M. succiniciproducens has been reported to produce succinic acid very efficiently (11). On the other hand, succinic acid is a minor product of mixed-acid fermentation in E. coli. It was reasoned that production of succinic acid in E. coli would be possibly enhanced by mimicking the metabolism of M. succiniciproducens. Toward this goal, we compared the central metabolic pathways of E. coli and M. succiniciproducens, aiming to identify candidate metabolic genes that possibly determine the different fermentation patterns in the two bacteria. Considering that the genome size of E. coli is about twice as large as that of M. succiniciproducens, we focused on metabolic pathways that are found in E. coli but not in M. succi-

4 VOL. 71, 2005 METABOLIC ENGINEERING FOR SUCCINIC ACID PRODUCTION 7883 niciproducens, since they were thought to drive metabolic flux away from succinic acid formation in E. coli. Compared in Fig. 1, succinate dehydrogenase (sdhabcd), malate dehydrogenase (mqo), glyoxylate shunt (aceba), and ptsg homologue genes were found in E. coli but not in M. succiniciproducens. While E. coli has two pyruvate kinases, A and F, encoded by the pyka and pykf genes, M. succiniciproducens has pyruvate kinase A only. Even though the sfca, fumab, poxb, and acs genes were only present in E. coli, they were not considered as knockout candidates for the following reasons. The malic enzyme encoded by the sfca gene has been reported to be beneficial for succinic acid production (12, 27). The fumarases encoded by the fumab genes do not seem to negatively affect succinic acid production because fuma expression is strongly repressed under anaerobic conditions (32), while FumB generates fumarate for use as an anaerobic electron acceptor (33). The products of the poxb and acs genes play important roles in efficient aerobic growth (1) and growth on acetic acid (20), respectively, and thus do not seem to affect succinic acid production under anaerobic conditions. Therefore, we selected five genes or operons, ptsg, pykf, mqo, sdhabcd, and aceba, as the initial targets to be inactivated to make E. coli mimic M. succiniciproducens. Comparative metabolic engineering. Comparative engineering of E. coli metabolic pathways was performed based on the above findings. The ptsg, pykf, mqo, sdhabcd, and aceba genes in E. coli, which were not found in the genome of M. succiniciproducens, were sequentially disrupted to examine their effects on fermentation profiles. The mutant strains constructed are listed in Table 1. First, the ptsg and pykf genes involved in pyruvate formation were inactivated, hoping that more PEP will be available for the carboxylation reaction. Then, the mqo and sdha genes were sequentially deleted to prevent the reverse flux from succinic acid. Finally, the aceba genes responsible for the glyoxylate shunt were inactivated. Anaerobic fermentation profiles of these mutant strains are summarized in Table 3. Interestingly, mutants rationally constructed by comparative metabolic engineering did not change the fermentation profiles much compared with the wild-type strain. Even the most heavily engineered W3110GFOHE strain, which was designed to be the strain most similar to M. succiniciproducens, did not show a significant difference in the growth rate and fermentation patterns compared with the wild-type strain (Table 3). Metabolic engineering based only on genome and pathway comparison was not successful, probably because we did not examine all possible combinational mutations of the candidate reactions. Since it is practically impossible to construct all these possible combinatorial mutant strains, in silico knockout experiments were carried out. In silico prediction of gene candidates for enhanced succinic acid production. The effects of combinatorial gene knockouts on succinic acid formation were examined in silico using the genome-scale E. coli metabolic model (21). The maximum biomass formation rate was used as an objective function during the optimization. When the enzymes (genes) were knocked out, the fluxes of the corresponding metabolic reactions were set to zero. It should also be noted that the metabolic reactions catalyzed by the pyruvate kinases A and F cannot be distin- e Not detected. a Anaerobic vial fermentation on 50 mm initial glucose for 24 or 80 h. b Residual glucose concentration. c Calculated as succinic acid/(succinic acid lactic acid formic acid acetic acid ethanol). d Calculated as succinic acid ratio/succinic acid ratio in the wild-type strain (0.017). Strains constructed based on in silico simulation resuts W3110GFA W3110GFA W3110GFAP ND e W3110GFAPL ND ND ND Strains constructed based on metabolic pathway comparison W3110G W3110GF W3110GFO W3110GFH W3110GFOH W3110GFOHE W Strain Fermentation time (h) Concn of substrate or product (mm) a Succinic acid Fold OD 600 Glucose b Succinic acid Lactic acid Formic acid Acetic acid Ethanol ratio c change d TABLE 3. Results of anaerobic fermentation of E. coli W3110 and its various mutant strains

5 7884 LEE ET AL. APPL. ENVIRON. MICROBIOL. In silico strain TABLE 4. Prediction of maximum biomass formation rate, succinic acid production rate, and succinic acid flux ratio in various in silico E. coli mutant strains Disrupted gene(s) Maximum biomass formation rate (h 1 ) Succinic acid production rate (mmole g dry cell weight 1 h 1 ) Succinic acid flux ratio a Wild type H sdha O mqo E aceba G ptsg FA pykfa HO sdha mqo HE sdha aceba HG sdha ptsg HFA sdha pykfa OE mqo aceba OG mqo ptsg OFA mqo pykfa EG aceba ptsg EFA aceba pykfa GFA ptsg pykfa HOE sdha mqo aceba HOG sdha mqo ptsg HOFA sdha mqo pykfa HEG sdha aceba ptsg HEFA sdha aceba pykfa HGFA sdha ptsg pykfa OEG mqo aceba ptsg OGFA mqo ptsg pykfa OEFA mqo aceba pykfa EGFA aceba ptsg pykfa HOEG sdha mqo aceba ptsg HOEFA sdha mqo aceba pykfa HOGFA sdha mqo ptsg pykfa HEGFA sdha aceba ptsg pykfa OEGFA mqo aceba ptsg pykfa HOEGFA sdha mqo aceba ptsg pykfa GFAP ptsg pykfa pfl GFAPL ptsg pykfa pfl ldha a Calculated as (succinic acid formation flux)/(summed formation fluxes of succinic acid, lactic acid, formic acid, acetic acid, and ethanol). guished in the in silico E. coli metabolic model because they catalyze the same reaction. The results of in silico experiments for the representative mutants are shown in Table 4. It was found that single-gene knockouts and many multiplegene knockouts did not significantly change the succinic acid production. The succinic acid production rate was increased by 100-fold when the ptsg and pykfa genes were knocked out in silico. Further knockout of the sdha, mqo, and aceba genes or their combinations did not increase the succinic acid production rate. However, inactivation of the pfl gene further increased the succinic acid production rate (Table 4). In conclusion, the in silico simulation results suggested that three pyruvate-forming enzymes encoded by the ptsg and pykfa genes need to be inactivated for the enhanced production of succinic acid in E. coli, which can be further increased by inactivating the pfl gene. Validation of in silico prediction. According to the in silico prediction, E. coli W3110GFA was constructed by knocking out the ptsg and pykfa genes. Anaerobic fermentation of W3110GFA for 24 h resulted in an about 3.4-fold increase in succinic acid formation (from less glucose consumed) and a significant reduction of other fermentation products compared with the wild-type strain (Table 3). The succinic acid ratio increased by more than eightfold using the mutant strain. In 80 h, E. coli W3110GFA was able to convert 50 mmol of glucose to 17.4 mmol of succinic acid, which is more than 7 times higher than that produced by wild-type and other mutant strains constructed based on pathway comparison. It can also be seen that the formation of other fermentation products was considerably reduced in W3110GFA, resulting in the 9.23-fold increase in succinic acid ratio compared with the wild-type strain (Table 3). These results suggest that the anaerobic fermentation pattern of E. coli could be dramatically affected by simultaneous disruption of strong pyruvate-forming enzymes present in E. coli as predicted by in silico analysis. It also indicates that blocking the major pathways converting PEP to pyruvate is beneficial in redirecting the metabolic flux to succinic acid in E. coli.

6 VOL. 71, 2005 METABOLIC ENGINEERING FOR SUCCINIC ACID PRODUCTION 7885 Strain TABLE 5. Effects of external supply of pyruvate on the anaerobic fermentation profile of E. coli W3110GFA Concn of initial carbon source (mm) OD 600 Concn of substrate or product (mm) a Succinic acid ratio c Glucose Pyruvate Glucose b Succinic acid Lactic acid Formic acid Acetic acid Ethanol W W3110GFA W3110GFA W3110GFA ND e a Anaerobic vial fermentation for 24 h. b Residual glucose concentration. c Calculated as succinic acid/(succinic acid lactic acid formic acid acetic acid ethanol). d Calculated as succinic acid ratio/succinic acid ratio in the wild-type strain (0.017). e Not detected. Foldd change Further disruption of the pfl and ldha genes in a W3110GFA strain. With an aim to further reduce other fermentation products, the pfl and ldha genes were inactivated in a W3110GFA strain. An E. coli W3110GFAP strain lacking the pfl gene showed slower growth than its parental W3110GFA strain. Formic acid was not detected, and acetic acid formation was dramatically decreased. However, a twofold increase of lactic acid formation was observed. Even though the final succinic acid concentration obtained with W3110GFAP was less than half of that obtained with W3110GFA, the succinic acid molar ratio was increased by 12.6-fold (Table 3), which is consistent with the in silico prediction results (Table 4). The W3110GFAPL strain, additionally lacking the ldh gene, showed normal growth under aerobic conditions but only marginal growth under anaerobic conditions. It could utilize only a little glucose under anaerobic conditions because of the lack of major anaerobic respiratory pathways (Table 3). In the absence of Na 2 S, slightly enhanced succinic acid formation ( 2 mm) was observed in the anaerobic culture of W3110GFAPL. These results suggest that pyruvate-forming enzymes (encoded by the ptsg, pykf, and pyka genes) and pyruvate-dissimilating enzymes (encoded by the ldha and pfl genes) play essential roles in succinic acid production by E. coli. Pyruvate supply affects the fermentation profiles of a W3110GFA strain. To examine the effects of pyruvate on the anaerobic metabolism of E. coli, pyruvate was supplied externally during the fermentation of E. coli W3110GFA. It was found that the addition of glucose plus pyruvate in the medium changed the active succinic acid fermentation pattern of W3110GFA to the mixed-acid fermentation, as in the wild-type strain (Table 5). When pyruvate was added as a sole carbon source, significant amounts of formic and acetic acids were formed. Notably, the succinic acid ratio also returned to that of the wild-type strain by the addition of pyruvate. These results suggest that pyruvate is mainly converted to formic and acetic acids rather than succinic acid in the anaerobic fermentation of E. coli. DISCUSSION E. coli carries out mixed-acid fermentation under anaerobic conditions and produces much more acetic acid, formic acid, lactic acid, and ethanol than succinic acid. In contrast, a rumen bacterium, M. succiniciproducens, can produce succinic acid as a major fermentation product. It was therefore assumed that different fermentation patterns might be caused by different fermentation pathways in operation in two bacteria. To make E. coli mimic the fermentation pattern of M. succiniciproducens, several E. coli mutants were constructed based on the results of pathway comparison (Table 1). The ptsg, pykf, mqo, sdhabcd, and aceba genes were selected as the target genes to be disrupted. It should be mentioned that M. succiniciproducens does not operate the PTS. We have previously reported that M. succiniciproducens transports glucose by the PTS if we accept the annotation results obtained using the clusters of orthologous group database. On the other hand, the annotation results based on the nonredundant database suggest that M. succiniciproducens does not utilize the PTS for glucose uptake. Therefore, we carried out actual experiments and found that the M. succiniciproducens does not utilize the PTS. E. coli W3110 was able to efficiently phosphorylate glucose using PEP or ATP as a phosphate donor. On the other hand, M. succiniciproducens could efficiently transfer phosphate from ATP to glucose but rather poorly from PEP. This is the reason why the E. coli ptsg gene was chosen as one of the candidate genes to be knocked out. However, even the most heavily engineered W3110GFHOE strain did not show significant changes in the fermentation profiles, suggesting that the robust metabolism aided by many shunts and isoenzymes present in E. coli hinders the flux changes. Thus, metabolic engineering based on simple yet rational comparison of the two genome-scale metabolic pathways was not successful. This was thought to be because we could not examine all the possible combinatorial knockout mutant strains. Since it is practically impossible to generate all the possible combinatorial mutant strains, we decided to carry out in silico knockout experiments. However, it requires great computational power and a long time to perform in silico simulations for all possible combinations. It was thus invaluable to have the results of rational genome comparison, as we could focus the simulations on those genes that were found to be important for succinic acid production. The results of in silico experiments suggested that knocking out the ptsg, pykf, and pyka genes is the most beneficial for the enhanced production of succinic acid. It is interesting to note that in silico E. coli cannot distinguish the products of the pykf and pyka genes, as they catalyze the same reaction. During the comparative metabolic engineering studies, the pyka gene was not considered a candidate gene to be knocked out because M. succiniciproducens also possessed this gene. It was, however, concluded from the in silico analysis that reducing

7 7886 LEE ET AL. APPL. ENVIRON. MICROBIOL. the pyruvate-forming flux is very important for enhanced succinic acid production. The actual experiments showed the predicted results. E. coli W3110GFA produced significantly more succinic acid and concomitantly less of other fermentation products, though it grew slightly slower than the wild-type strain under anaerobic conditions (Table 3). It is important to note that the metabolic fluxes could be altered to lead to enhanced succinic acid formation by manipulating the upstream pathways (e.g., the ptsg, pykf, and pyka genes) rather than the pathways directly involved in the end product formation (such as the ldha, pfl, and pta genes). These findings are reasonable, as high-level succinic acid producers such as M. succiniciproducens and A. succinogenes efficiently convert PEP to succinic acid via oxaloacetate, malate, and fumarate (11, 29). Therefore, the metabolic flux to pyruvate is not beneficial for succinic acid production, since pyruvate is efficiently converted to other fermentation products in E. coli. In a pyruvate complementation test, the formation of acetic and formic acids was significantly increased by the addition of pyruvate in the pyruvate-deficient W3110GFA strain (Table 5). Apparently, pyruvate is more readily converted to acetic and formic acids in E. coli. While E. coli has strong pathways for converting pyruvate ultimately to acetic acid (encoded by the poxb, pta, acka, and acs genes) to generate ATP, M. succiniciproducens has relatively weak pyruvate-forming pathways, since it has no active PTS for glucose uptake and only one pyruvate kinase (compared with two in E. coli) as mentioned above. This is the reason why M. succiniciproducens can efficiently produce succinic acid via anaplerotic pathways while sustaining the low level of pyruvate. Therefore, pyruvate was found to be the most important central metabolite affecting, most significantly, the fermentation pattern of E. coli. This difference seems to determine the characteristic fermentation patterns of the two microorganisms. In summary, successful metabolic engineering of E. coli for enhanced succinic acid production could be achieved by combining genome and pathway comparison, in silico metabolic characterization, and validation by knockout experiments. This strategy may be generally useful in improving microbial strains for the enhanced production of metabolites. ACKNOWLEDGMENTS We thank B. L. Wanner for providing plasmid pkd46 and E. Dervyn for plasmids pic156 and puc19-phleo. This work was supported by the Genome-Based Integrated Bioprocess Project of the Ministry of Science and Technology and by the BK21 project. Further support by the LG Chem Chair Professorship, IBM-SUR program, Microsoft, and the Center for Ultramicrochemical Process Systems sponsored by KOSEF is appreciated. REFERENCES 1. Abdel-Hamid, A. M., M. M. Attwood, and J. R. Guest Pyruvate oxidase contributes to the aerobic growth efficiency of Escherichia coli. Microbiology 147: Bunch, P. K., F. Mat-Jan, N. Lee, and D. P. Clark The ldha gene encoding the fermentative lactate dehydrogenase of Escherichia coli. Microbiology 143: Chao, Y. P., and J. C. Liao Alteration of growth yield by overexpression of phosphoenolpyruvate carboxylase and phosphoenolpyruvate carboxykinase in Escherichia coli. Appl. Environ. Microbiol. 59: Chatterjee, R., C. S. Millard, K. Champion, D. P. Clark, and M. I. Donnelly Mutation of the ptsg gene results in increased production of succinate in fermentation of glucose by Escherichia coli. Appl. Environ. Microbiol. 67: Datsenko, K. A., and B. L. Wanner One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc. Natl. Acad. Sci. USA 97: Davis, C. P., D. Cleven, J. Brown, and E. Balish Anaerobiospirillum, a new genus of spiral-shaped bacteria. Int. J. Syst. Bacteriol. 26: Dervyn, E., C. Suski, R. Daniel, C. Bruand, J. Chapuis, J. Errington, L. Janniere, and S. D. Ehrlich Two essential DNA polymerases at the bacterial replication fork. Science 294: Edwards, J. S., and B. O. Palsson The Escherichia coli MG1655 in silico metabolic genotype: its definition, characteristics, and capabilities. Proc. Natl. Acad. Sci. USA 97: Gokarn, R. R., J. D. Evans, J. R. Walker, S. A. Martin, M. A. Eiteman, and E. Altman The physiological effects and metabolic alterations caused by the expression of Rhizobium etli pyruvate carboxylase in Escherichia coli. Appl. Microbiol. Biotechnol. 56: Guettler, M. V., M. K. Jain, and B. K. Soni. April Process for making succinic acid, microorganisms for use in the process and methods of obtaining the microorganisms. U.S. patent 5,504, Hong, S. H., J. S. Kim, S. Y. Lee, Y. H. In, S. S. Choi, J. K. Rih, C. H. Kim, H. Jeong, C. G. Hur, and J. J. Kim The genome sequence of the capnophilic rumen bacterium Mannheimia succiniciproducens. Nat. Biotechnol. 22: Hong, S. H., and S. Y. Lee Metabolic flux analysis for succinic acid production by recombinant Escherichia coli with amplified malic enzyme activity. Biotechnol. Bioeng. 74: Hou, B. K., J. S. Kim, J. H. Jun, D. Y. Lee, Y. W. Kim, S. Chae, M. Roh, Y. H. In, and S. Y. Lee BioSilico: an integrated metabolic database system. Bioinformatics 20: Kim, P., M. Laivenieks, C. Vieille, and J. G. Zeikus Effect of overexpression of Actinobacillus succinogenes phosphoenolpyruvate carboxykinase on succinate production in Escherichia coli. Appl. Environ. Microbiol. 70: Lee, D. Y., H. Yun, S. Park, and S. Y. Lee MetaFluxNet: the management of metabolic reaction information and quantitative metabolic flux analysis. Bioinformatics 19: Lee, P. C., S. Y. Lee, S. H. Hong, and H. N. Chang Isolation and characterization of a new succinic acid-producing bacterium, Mannheimia succiniciproducens MBEL55E, from bovine rumen. Appl. Microbiol. Biotechnol. 58: Lin, H., G. N. Bennett, and K. Y. San Metabolic engineering of aerobic succinate production systems in Escherichia coli to improve process productivity and achieve the maximum theoretical succinate yield. Metab. Eng. 7: Millard, C. S., Y. P. Chao, J. C. Liao, and M. I. Donnelly Enhanced production of succinic acid by overexpression of phosphoenolpyruvate carboxylase in Escherichia coli. Appl. Environ. Microbiol. 62: Neidhardt, F. C., and H. E. Umbarger Chemical composition of Escherichia coli, p In F. C. Neidhardt, R. Curtiss III, J. L. Ingraham, E. C. C. Lin, K. B. Low, B. Magasanik, W. S. Reznikoff, M. Riley, M. Schaechter, and H. E. Umbarger (ed.), Escherichia coli and Salmonella: cellular and molecular biology, 2nd ed., vol. 1. ASM Press, Washington, D.C. 20. Oh, M. K., L. Rohlin, K. C. Kao, and J. C. Liao Global expression profiling of acetate-grown Escherichia coli. J. Biol. Chem. 277: Reed, J. L., T. D. Vo, C. H. Schilling, and B. O. Palsson An expanded genome-scale model of Escherichia coli K-12 (ijr904 GSM/GPR). Genome Biol. 4:R Sambrook, J., and D. W. Russell Molecular cloning: a laboratory manual, 3rd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. 23. Sanchez, A. M., G. N. Bennett, and K. Y. San Novel pathway engineering design of the anaerobic central metabolic pathway in Escherichia coli to increase succinate yield and productivity. Metab. Eng. 7: Sauer, U., D. C. Cameron, and J. E. Bailey Metabolic capacity of Bacillus subtilis for the production of purine nucleosides, riboflavin, and folic acid. Biotechnol. Bioeng. 59: Seok, Y. J., P. P. Zhu, B. M. Koo, and A. Peterkofsky Autophosphorylation of enzyme I of the Escherichia coli phosphoenolpyruvate:sugar phosphotransferase system requires dimerization. Biochem. Biophys. Res. Commun. 250: Steinmetz, M., and R. Richter Plasmids designed to alter the antibiotic resistance expressed by insertion mutations in Bacillus subtilis, through in vivo recombination. Gene 142: Stols, L., and M. Donnelly Production of succinic acid through overexpression of NAD -dependent malic enzyme in an Escherichia coli mutant. Appl. Environ. Microbiol. 63: Stols, L., G. Kulkarni, B. G. Harris, and M. I. Donnelly Expression of Ascaris suum malic enzyme in a mutant Escherichia coli allows production of succinic acid from glucose. Appl. Biochem. Biotechnol : Van der Werf, M. J., M. V. Guettler, M. K. Jain, and J. G. Zeikus Environmental and physiological factors affecting the succinate product ratio

8 VOL. 71, 2005 METABOLIC ENGINEERING FOR SUCCINIC ACID PRODUCTION 7887 during carbohydrate fermentation by Actinobacillus sp. 130Z. Arch. Microbiol. 167: Varma, A., and B. O. Palsson Stoichiometric flux balance models quantitatively predict growth and metabolic by-product secretion in wildtype Escherichia coli W3110. Appl. Environ. Microbiol. 60: Vemuri, G. N., M. A. Eiteman, and E. Altman Effects of growth mode and pyruvate carboxylase on succinic acid production by metabolically engineered strains of Escherichia coli. Appl. Environ. Microbiol. 68: Woods, S. A., and J. R. Guest Differential roles of the Escherichia coli fumarases and fnr-dependent expression of fumarase B and aspartase. FEMS Microbiol. Lett. 48: Woods, S. A., S. D. Schwartzbach, and J. R. Guest Two biochemically distinct classes of fumarase in Escherichia coli. Biochim. Biophys. Acta 954: Zeikus, J. G., M. K. Jain, and P. Elankovan Biotechnology of succinic acid production and markets for derived industrial products. Appl. Microbiol. Biotechnol. 51: